You're assuming that there is an obvious,
rationale, easily explainable reason why
performance of a suboptimal implementation
works better on one box than another.
The reason, while unknown to me, is likely
due to a complex, non-obvious problem.
It could have something to do w/ mesh-ed-ness
of scope of external routes.
Again, I dunno, I just run IS-IS and don't
have to deal with these issues.
-alan
Thus spake Sean Butler (sbutler1@tampabay.rr.com)
on or about Thu, Oct 28, 1999 at 01:02:42PM -0400:
>
> Alan Hannan wrote:
> >
> > It is a generally accepted perception that
> > cisco OSPF does not scale past a certain point,
> > and that external routes in OSPF exacerbate the
> > overhead issue quite significantly.
> >
> > I'd highly recommend you transition to IS-IS.
> >
>
> This would not really explain why SPF calc's run
> longer on about 3 out of a 70 or so routers that
> are all in the same area, when these 3 routers are
> *not* the most heavily loaded in terms of packets
> per second, bits per second, interfaces, bgp
> peers, ospf neighbors, frame pvc's, etc. etc...
>
> There are less than 2000 external routes in OSPF,
> and while that is high, the network was running
> with 6000 until we did some iBGP things, without
> exhibiting this strange behaviour.
>
> Add to that the fact that SPF calc. triples with
> CEF enabled on these 3 boxes only, and it makes
> for something extremely odd!
>
> Thanks!
> /Sean
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:07 EDT